Is Perception a Source of Reasons?theo_1139

Transcription

1 bs_bs_banner THEORIA, 2013, 79, doi: /j x Is Perception a Source of Reasons?theo_1139 by SANTIAGO ECHEVERRI University of Geneva Abstract: It is widely assumed that perception is a source of reasons (SR). There is a weak sense in which this claim is trivially true: even if one characterizes perception in purely causal terms, perceptual beliefs originate from the mind s interaction with the world. When philosophers argue for (SR), however, they have a stronger view in mind: they claim that perception provides pre- or non-doxastic reasons for belief. In this article I examine some ways of developing this view and criticize them. I exploit these results to formulate a series of constraints that a satisfactory account of the epistemic role of perception should fulfil. I also make a positive suggestion: coherentists are right when they claim that only beliefs can be reasons for other beliefs. Nevertheless, I depart from traditional coherentism, for I do not buy its conception of perception as bare sensation, nor explicate the justificatory status of beliefs in terms of coherence. My point is rather that, when one invokes experience to justify a belief, the justifying state must have structural features of beliefs. Keywords: epistemology of perception, belief, reasons, belief revision SUPPOSE YOU ACCEPT A WIDESPREAD analysis according to which mental acts or states like believing, desiring, thinking, or intending have a three-place structure: a subject S is related via a psychological act or state y to a propositional content p: S y s that p. Suppose further that you think, as most contemporary philosophers do, that perception grounds some mental acts or states. These two assumptions should lead you to ask how perception is related to the attitudes listed above. An influential response says that perception is a source of reasons (hereafter SR ). This suggestion should satisfy two demands. First, one has to provide an account of reasons. Second, it is necessary to show that, given one s preferred ontology of reasons, perceptual states adequately ground some beliefs. Concerning the first point, there are two conceptions of reasons in the literature: they are seen either as structured contents, like facts or propositions, or as mental states, like beliefs. One s answer to the second question is a bit tricky. Suppose someone conceives of perception in purely causal terms. In that case, perceptual beliefs are grounded in perception because the latter triggers them. Some philosophers have complained, however, that bare causal relations are too weak. As a result, they have argued for a picture in which perception also plays a normative role. In this article I will criticize normative versions of (SR): the strong propositional view (sections 2 4), the weak propositional view (section 5), and the Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA.

2 IS PERCEPTION A SOURCE OF REASONS? 23 quasi-doxastic view (sections 6 7). My main purpose is to show that none of these accounts provides a satisfactory picture of the way beliefs are grounded in perception. I shall use these results to formulate a series of constraints that a satisfactory account of the epistemic role of perception should fulfil. I will also suggest that, strictly speaking, only beliefs can normatively ground other beliefs. This might be seen as a defence of coherentism. Nevertheless, this labelling is misleading for two reasons: first, recent versions of coherentism introduce nondoxastic justifiers (Kvanvig, 2011; Gupta, 2006a) and, second, coherentism is associated with the controversial claim that justification supervenes on the internal coherence of a belief system. My purpose is rather to motivate a new epistemological model of perception in which the latter is not conceived as a reason for belief. 1. Structural vs. Normative Adequacy (SR) is the claim that experience delivers pre- or non-doxastic reasons for belief. The main motivation for this view is to show that experience can play a rational role. If this picture is right, most coherentists are wrong to claim that only beliefs can be invoked as reasons for other beliefs. Thus, despite its pre- or non-doxastic character, perceptual experience provides an adequate rational basis for perceptual beliefs. What does rational adequacy mean? We can distinguish two different notions: structural and normative. An entity E is structurally adequate to justify a belief if E satisfies two conditions: (1) it has features allowing it to participate in rational relations, and (2) it can be exploited to provide an account of the way perceptual beliefs are based or grounded. In this context, I define basing (or grounding ) in explanatory terms. A subject s belief B is based on E just in case she holds B because of E. Intuitively, chairs or hurricanes are structurally inadequate because they lack the features enabling them to participate in rational relations. Moreover, it is difficult to understand how the bare existence of any of those entities could fully explain one s perceptual belief that there is a chair or that a hurricane is coming. The question of structural adequacy is to be distinguished from that of normative adequacy. The latter is related to the factors that positively or negatively affect the normative standing of propositions or beliefs (Goldman, 2009; Pryor, 2005). In this sense, a normative account of the epistemic role of perception purports to explain which factors make perceptual beliefs justified, reasonable or instances of knowledge. As a result, providing a structural account of the epistemic role of perception is orthogonal to the project of delivering a normative account. After all, factors that are external to the perception-belief relation may affect the normative standing of

3 24 SANTIAGO ECHEVERRI a perceptual belief. That the visual system be in good order or that the illumination conditions be normal are factors that affect the normative standing of visual beliefs. Nevertheless, they are not directly relevant to the structural account of the way visual beliefs are grounded in perception. A consequence of the distinction between structural and normative adequacy is that one can investigate the connection between perception and belief without taking a definitive stance on the debate between internalism and externalism. This debate is related to the question whether (most of) the factors that positively or negatively affect the normative standing of propositions or beliefs are internal or external. Although important, this question is different from the structural issue. The latter is a problem for any theorist who agrees that some beliefs are grounded in perception, or accepts that perceptual experience can be cited in justificatory practices. Thus, it is possible to combine various structural accounts of the perception-belief relation with externalist accounts in terms of factive evidence, reliable processes or teleological functions. 1 Coherentism is typically cited as a rejection of (SR). Given the above distinction, however, this example is misleading, for it combines normative and structural considerations. According to a common characterization, coherentism argues that the factors that positively or negatively affect the normative standing of beliefs are (mostly) internal to belief systems. Interestingly, this normative claim has been motivated on the structural assumption that a direct confrontation of beliefs with the world makes little sense. This gives rise to Davidson s (1986) famous contention that perception plays no evidential role in the justification of beliefs, for only a belief may count as a reason for holding another belief. The claim has an ontological scope: perceptual states are not the sorts of things that may stand in rational relations to beliefs. Davidson s view follows from an assimilation of perceptual states to sensations, which only bear a causal relation to propositional attitudes. The claim can be put as follows: although reasons may be causes, not all causal relations are rational. According to Davidson, given that perceptual states only bear a causal relation to beliefs, they do not provide reasons for belief. But this view might be justified on different considerations. If one s ontology prohibits the postulation of structured entities like facts, states of affairs or propositions, one may be led to endorse Davidson s view (see Turri, 200, pp and n. 5). A different approach would be to look at perceptual experience in order to see whether it bears a rational relation to beliefs. This is the path I will explore here. In this article, I will argue for what Lyons (2009) terms the Belief Principle (BP): only beliefs can evidentially justify other beliefs. Hence, my agreement with 1 I develop a functional-teleological account of the normative issue in Echeverri (submitted).

4 IS PERCEPTION A SOURCE OF REASONS? 25 Davidson is partial. Whereas Davidson defined the coherentist view as a defence of this structural claim, I will try to develop it independently of his more controversial views on the factors that make a belief justified. Furthermore, I will show that one can share Davidson s reliance on (BP) without buying his conception of perception as bare sensation. Providing new arguments for (BP) is important because (SR) has become the orthodoxy in many circles. 2 Since defenders of (SR) have tried to provide an alternative to (BP), they are committed to showing that other entities can participate in rational relations, and explain how perceptual beliefs are grounded in perceptual states. In other words, they have to demonstrate that some pre- or non-doxastic states are structurally adequate. By looking at things in this light, the question is whether pre- or nondoxastic propositional states are structurally adequate, not whether they are normatively adequate. The latter issue would lead us to include an analysis of non-evidential normative factors as well. These issues are not my main concern in this article. 3 My overall strategy will be as follows. I will examine three versions of (SR), and show that none of them provides an adequate account of the basing relation. My objections will be of two sorts: psychological and phenomenological. I will exploit the results to formulate a series of constraints that should be satisfied in order to get the basing relation right. Since the motivation of these constraints will follow from the discussion, I will not try to back them up by independent arguments. If the reader thinks she could resist them, she ought to find a flaw in the discussion that motivates them. 2 An important exception is Jack Lyons. I got to know Lyons work thanks to a referee of this journal. Although my picture is congenial to his view in some respects, our motivations are different. Whereas Lyons thinks that qualia are epistemologically irrelevant, I motivate my rejection of (SR) on the examination of three propositional accounts of experience. Contrary to Lyons, I do not understand phenomenology as an endorsement of qualia but as a descriptive discipline that investigates the way things appear. Thus, one can use phenomenology to draw some philosophical conclusions without positing qualia, which is more controversial. Whereas Lyons puts his view in the service of reliabilism, I emphasize a teleological analysis. For further details, see Echeverri (submitted). 3 My distinction between structural and normative adequacy differs from Ginsborg s (2006) suggestion that Davidson s claim concerns one sense of the notion of a reason, what she calls third-person reasons, as opposed to first-person reasons. She thinks Davidson s view is restricted to the use of beliefs to rationalize someone s behaviour from the third-person point of view (Ginsborg, 2006, p. 302). Although I cannot examine this picture here, I think Ginsborg is wrong to smuggle facts into the characterization of first-person reasons. This has an unfortunate consequence. If one invokes facts as part of the definition of reasons in the first-person sense, the question arises as to what entity might work as a reason for belief when there are no available facts. Ginsborg (2006, pp , n. 10) acknowledges this case, and introduces supposed facts (p. 303) to accommodate perceptual error. Nonetheless, it is unobvious what it means to be presented with a supposed fact. For further discussion, see Turri (2009, p. 502) and Echeverri (2011). I am sympathetic, however, to Ginsborg s (2006, pp ) critical remarks on non-doxastic conceptions of experience, as the rest of the article shows.

5 26 SANTIAGO ECHEVERRI 2. The Strong Propositional View A number of philosophers think that perceptual states have propositional content. An example of this view is conceptualism, i.e., the claim that the content of perceptual experience is identical to the content expressed by an embedded that - clause preceded by a cognitive verb (McDowell, 1994a, 1998a; Brewer 1999). Conceptualists think this claim entails that experiences are conceptually articulated. This follows from their assumption that the building blocks of propositions are concepts. This conclusion is not mandatory, though. One could hold that the building blocks of propositions are individuals and properties (see Kaplan, 1977). If that is right, it should be possible to develop the propositional view in a nonconceptualist guise. For reasons of space, I will bracket the debate about the nature of propositions, and focus on the general assumption that they are structured entities. For a content to count as propositional in this minimal sense it must have a compositional structure that can be assigned a Boolean value: True or False. Although there are accounts of propositions as sets of possible worlds (e.g., Stalnaker, 1984), the view of propositions as structured entities has been more influential within studies of perception. 4 Does this account of the content of perception provide an adequate model of (SR)? If one holds that reasons are structured entities, like facts or propositions, one might be inclined to answer yes. If one grants that perception has structured content, one has solved the main problem of (SR). As I will explain later, however, this move is too hasty, for it fails to address the basing problem. To reach this conclusion, I will make use of the writings of conceptualists because they provide a detailed version of the propositional view. Nevertheless, the arguments do not depend on the conceptual or non-conceptual character of experience. Thus, we can extend them to propositional views that are either hostile to conceptualism (e.g., Peacocke, 1992, 2001) or neutral on the nature of perceptual content (e.g., Chudnoff, 2010; Huemer, 2001, 2007). In their earlier writings, John McDowell and Bill Brewer grant that experience has propositional content, and analyse it as having constituent structure. At the same time, they declare that perceiving is structured before the subject forms a judgement or fixes a belief. 5 McDowell declares: 4 Some of the points I make in this article can be translated into the framework of possible-worlds semantics. 5 For the purposes of this article, I take beliefs as mental dispositions and judgements as mental acts. From this perspective, a belief Bp may be manifested by different judgements J 1p, J 2p...J np. I remain neutral on whether judgements are conceptually prior to beliefs or vice versa.

6 IS PERCEPTION A SOURCE OF REASONS? 27 That things are thus and so is the content of the experience, and it can also be the content of a judgment. It becomes the content of a judgment if the subject decides to take the experience at face value. (McDowell, 1994a, p. 26; see also p. 62 and his 1998a, pp ) Brewer is equally explicit on this point: [T]he move from unendorsed understanding of these perceptual demonstrative contents [delivered by perceptions] to endorsement of them in belief will normally be almost instantaneous, and only rarely involve anything like a considered decision on the subject s part. They are importantly distinct stages of the story, though. The first is something for which the subject need have no epistemically relevant reasons, which in turn provides her with reasons of a genuinely epistemic kind for the second. (Brewer, 1999, p. 223 n. 5; emphasis added) 6 A distinctive feature of this approach is the claim that perceptual experience itself is structured. One s experience is structured as p before one endorses p or takes it to be p. This contrast is made vivid by the use of different verbs, which are supposed to denote different psychological stages: first, there is a grasping or understanding of p in perceptual experience; later on, one accepts or endorses that content in a judgement or belief. 7 From a classical viewpoint, having a perceptual experience is like entertaining a proposition, whereas judging or believing would introduce a propositional attitude in its own right. 8 What are the motivations for this propositional view? We can indicate two: Implausibility of Doxastic Accounts: At the time McDowell and Brewer developed this picture, it was widely assumed that structured contents were only available to characterize cognitive states like beliefs. Yet McDowell and Brewer also assumed, as most contemporary theorists do, that experiences are independent of belief. Given that doxastic accounts of experience were thought to be implausible, the claim that experience has propositional content led to the view that it is a non-doxastic state. In the early formulation, perceptual experience is a pre-doxastic state, i.e., a state that precedes perceptual judgement or belief. Experience is a pre-doxastic propositional state that provides reasons for belief. A cursory examination of recent proposals shows that this view remains influential (see Huemer, 2007, pp ; Chudnoff, 2010). Normativity: The conceptualist view is driven by an additional concern: to show that the mind s perceptual relation to the world is intrinsically normative even 6 For a more recent defence of conceptualism, see Brewer (2005). Nevertheless, Brewer (2006) criticizes the claim that experience has correctness conditions and argues for a non-representational theory. 7 I shall ignore the contrast between acceptance and belief. Although important, it is orthogonal to the arguments in this article, and it does not play any role in the theories of experience considered here. 8 According to Smith (2001), the view of perceptual experience as the entertaining of a proposition was already developed by Runzo (1977, pp ): To perceive an object or state of affairs, X, is, and is no more than, to be episodically aware of a set of propositions about X [...] This awareness of (sets of) propositions during perceiving is akin to entertaining.

7 28 SANTIAGO ECHEVERRI though it is not to be conceived in the model of belief. Normativity is one of the main tenets of McDowell s conceptualism: he is convinced that only a normative account of perception would avoid what Sellars termed the Myth of the Given, and provide an alternative account to Davidson s coherentism. Since the interpretation of the colourful expression Myth of the Given is not easy, I will leave it for another occasion. For the time being, it is important to bear in mind that the propositional view is driven by the aim of conceiving the mind s relation to the world in normative terms. One of the driving forces of the view is to make room for a conception of experience that can work as a tribunal, i.e., as a test for beliefs (for discussion, see Gupta, 2006b). The prior considerations suggest that propositional accounts are committed to the bold claim that pre-doxastic propositional states are sufficient to ground perceptual beliefs. For this reason, I will call it the strong propositional view. Let us take a look at the conceptualist version of this picture. Conceptualists identify the content of perceptual judgements with the content of experiences. They locate perception within the space of reasons but preserve the distinction between inferential and non-inferential sources of knowledge. They introduce a non-inferential but still rational perception-judgement transition. The link proceeds by sameness of content. This idea can be clarified by comparing this move with instances of the repetition rule: p, p. This rule asserts the antecedent and asserts the consequent. If we use Frege s assertion sign to mark the assertive force, we could symbolize the repetition rule thus: p, p. By contrast, according to our reconstruction, conceptualism introduces a transition from an unendorsed to an endorsed content, something like this: p, p. In this view, although the same content appears both in the antecedent and the consequent, it does not instantiate the repetition rule. This reconstruction enables the strong propositionalist to accommodate the defeasible character of perception-judgement transitions. If they merely happened to instantiate the repetition rule, it would be hard to understand how perception may lead to false belief. But if judgement leads to endorsement, one can try to make room for error as arising in the act of endorsement (see McDowell, 1998b, p. 405). 9 Additionally, this picture promises to capture an essential aspect of reasons, since it conceives of them as structured. Most contemporary theorists assume that being structured is a necessary condition for being part of our epistemic practices of deduction, probabilistic reasoning, induction, etc. 10 Besides, one might acknowledge with Sellars (1956), Brandom (1997), Williamson (2000), Steup (2001) and Pryor (2007), among others, that entities devoid of syntactic structure could not play the role of reasons: 9 For further discussion, see Glüer (2009) and Echeverri (2011). 10 Brewer (1999, pp ) introduces this argument. See also McDowell (1994a, p. 7).

8 IS PERCEPTION A SOURCE OF REASONS? 29 [O]nly things with sentential structure can be premises of inference [...]. For this reason sensings, understood in terms of nonepistemic relations between sense contents and perceivers, are not well suited to serve as the ultimate ground to which inferentially inherited justification traces back. (Brandom, 1997, p. 128; see also Brewer, 1999, p. 151) I shall call it the argument from structural specificity. It shows that only entities with syntactic complexity are apt to participate in justification relations. Let us illustrate its merits with an example taken from Williamson (2000, p. 195). Someone in court accused of murder could hardly be declared guilty just by presentation of a bloody knife to the judge. Doing this would be too unspecific to convict her. In order to decide whether she is guilty, it would be necessary to formulate propositions about the bloody knife, such as the knife was found in her house or, pointing to the accused, the knife has her fingerprints. Merely presenting the knife could hardly provide a reason to condemn her. As Williamson (2000, p. 195) points out, it is a source of indefinitely many such propositions. The argument stresses that rational relations are specific in a way in which entities lacking syntactic complexity are not. In some accounts, propositions have that complexity. Hence, in those accounts, if a state has propositional content, it can participate in rational relations. Since the strong propositional view is based on the conception of propositions as structured entities, it can explain how perception participates in rational relations. 3. The Strong Propositional View is False In the last section, I presented the strong propositional version of (SR). According to this view, perceptual experiences are pre-doxastic propositional states that are adequate to ground beliefs. Thus, the strong propositional view provides a genuine alternative to (BP). It conceives of perception-judgement transitions as moves from unendorsed to endorsed contents. In this section, I shall criticize this view on the basis of two arguments. As we shall see, these objections are neutral on the nature of reasons and justification, and they do not appeal to Davidson s conception of perception as bare sensation. First of all, the argument from structural complexity does not show that the content of perceptual experiences itself is propositional. What it establishes is the weaker claim that the evidence invoked in justificatory practices is propositional, not that experiences are propositionally structured. One might therefore argue that a non-propositional theory can accommodate the requirement of propositionality. Take transformation models of perception, where the transition from perception to belief involves a constructive process or a change in representational format. One can argue that propositions intervening in empirical justifications are built at the

9 30 SANTIAGO ECHEVERRI level of judgement, not at the level of pre-doxastic perceptual content (see, e.g., Burge, 2010). This proposal gains plausibility when one considers some experiences one would not intuitively classify as having propositional content, like the state of pain. Intuitively, I can explain to you that I am moaning because I am feeling pain. Here my state of pain is a source of justification. But it does not seem to have propositional structure. Similar remarks could be made against pictures that are neutral on the conceptual character of experience. Huemer (2001, 2007) and Chudnoff (2010) assume that grounding states must be propositional. The former defines his principle of phenomenal conservatism as follows: If it seems to S that p, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p (Huemer, 2007, p. 30). This family of views is caught on the horns of a dilemma: either they try to extend their account to pain or they restrict it to propositional states. In the former case, they face a serious problem: to give independent arguments for an intentionalist analysis of pain. Unfortunately, these arguments have not been provided by defenders of the strong propositional model. Thus, in the best possible scenario, strong propositional accounts have limited application. If they have limited application, however, one may wonder why one could not generalize from the example of pain to the perceptual case, and treat all cases as involving non-propositional grounds. The prior remarks show that it is not necessary to conceive of experiences as propositionally structured in order to satisfy the argument from structural specificity. What is needed is a model where propositions become available for justification, i.e., an account wherein subjects smoothly move from perceptual states to the formation of propositional states in their justificatory practices. We can exploit this remark to formulate the first constraint: Constraint 1: Any epistemological theory of perception must explain how specific states become available in experience. If the account posits nonpropositional states, it should bridge the gap between the non-propositional character of experience, and the propositional character of reasons. Second, although there are good reasons to think that propositionality is a necessary condition for reasons, having a propositional structure is not sufficient for a state to ground a perceptual belief. Let me explain. A number of propositional attitudes are said to have structured contents, although they do not play the role of epistemic reasons. Imagine that an opinionated subject is presented with the Müller-Lyer figure for the first time in her life. After being informed that it is an illusion, she does not believe it. The experimenter, somewhat surprised, insists: The two lines are of the same length. Line B [with the external hashes] merely looks longer than line A [with the internal hashes]. Why do you think B is longer

10 IS PERCEPTION A SOURCE OF REASONS? 31 than A? One can imagine a series of answers expressing structured states but which would yield inappropriate justifications: Wants: *Because I want line B to be longer than line A. Assumptions: *Because I assume that line B is longer than line A. Wonderings: *Because I wonder whether line B is longer than line A. These examples suggest a critique of the strong propositional view. In the attempt to provide a pre-doxastic picture of experience, it has severed an ingredient of justifying states: what Sellars (1956) aptly termed their endorsement dimension. The intuition behind this objection is that having a propositional structure is not sufficient to ground a belief. Grounding states are not just made out of propositions but also from what Frege called assertive force. Assertive force is constitutive of judgements and beliefs but, since the propositionalist wants to provide an alternative to doxastic theories of perception and resist (BP), she has lost this dimension from the picture. As indicated above, this argument can be applied to nonconceptualist views like the one favoured by Peacocke (1992, 2001). If one takes it that experiences deliver unendorsed non-conceptual propositional contents, and assumes that they ground some beliefs, one could be criticized on the same count. Propositionality is not (structurally) sufficient for a state to ground another state. And it can also be applied against any picture that assumes that perceptual reasons are pre-doxastic propositional states (Huemer, 2001, 2007; Chudnoff, 2010). This line of argument becomes more perspicuous when one considers the metaphor of the tribunal of experience, which conceptualists take as a test for any theory of perception. Contrary to Davidson, McDowell thinks this image is crucial to an understanding of human rationality because it allows for a characterization of belief systems as rationally sensitive to the impact of the world. According to McDowell, only a theory capable of making sense of perceptual experience as a source of reasons could make sense of that revealing metaphor. In the light of the previous remarks, it is clear that the strong propositionalist is not entitled to conceive of experience as a tribunal either. Just take the following example. My father and I plan to go and play tennis in the afternoon. In the meantime, I hear on the radio that it has been raining all day long and, as a result, I change my mind: Since it s raining, we won t play tennis. When my father comes, I greet him with the following comment: It s raining! It s impossible to play tennis! My father, who just came in from outdoors, replies: No, it isn t raining anymore. Are you ready to take revenge? How should we conceive of my process of belief revision? If I were a bit sceptic, I would probably walk to the nearest window in order to check if it is raining. If I were to see that the sun is shining, I ought to update my previous belief in order to match it to the present weather. It seems crucial, however, that I accept that the sun is shining. The fact that the sun is shining could hardly eventuate in a belief revision process if I were not

11 32 SANTIAGO ECHEVERRI to take it (at least tacitly) that the sun is shining. Only accepted contents can be considered in belief revision. Non-accepted contents would be ignored; after all, they do not tell me that the sun is shining. 11 One can make the same point by reflecting on the nature of Frege s (1879) assertion sign. In the concept-script, unendorsed contents never occur in isolation but only as constituents of more complex judged contents. This is what happens in conditionals. In a conditional of the form p, q, the antecedent figures as an entertained content. Yet, it is also a constituent of an asserted conditional. The motivation for this view is that only asserted contents can yield immediate justification. If an unendorsed content is invoked as a reason, it must be paired with an endorsed content before it leads to a conclusion. If one concludes q from (p, q) and p, one combines the assertion of (p, q) with the assertion of p. The same occurs when one appeals to the disjunctive syllogism: one combines the assertion of (pvq) with the assertion of q in order to conclude the assertion of p. Although p and q are entertained, they are not entertained in isolation but in the context of a disjunctive assertion. The strong propositional view of perception leads us to do violence to these facts: it assumes that entertained contents occur in isolation, and are sufficient to immediately ground perceptual beliefs. This contradicts our justificatory practices. 12 Before I draw the moral from this argument, let me clarify three points: First, the analogy with Frege s assertion sign shows that we are not advocating any form of Cartesianism. The point is not that a conscious appraisal of perceptual content is structurally necessary for belief revision. The claim is rather that merely entertained contents are psychologically idle. If one is interested in avoiding any Cartesian commitments, one can construe the notion of doxastic force in functional terms: in order to generate a belief revision process, the content must be appropriately linked to other beliefs. The intuition here is that merely entertained contents are not appropriately linked to other beliefs, so they cannot generate rational moves I am echoing Frege: How does a thought act? By being grasped and taken to be true (Frege, , p. 61; 1997, p. 344). 12 It might be objected that, when one reasons by reductio ad absurdum, one entertains an isolated proposition for the sake of the argument. Two replies are available here. First, this technique does not provide immediate justification in the sense intended by (SR). One proceeds by reductio when immediate justification is not available. Second, it is unclear that these cases involve entertaining a proposition, instead of some positive attitudes, like supposing or assuming it to be true. 13 The tribunal scenario is philosophically interesting because it casts doubt on the strong separation between action and belief. It recreates a situation in which one s epistemic reasons must generate an adequate epistemic action on the subject s side: to update her belief system. Thus, it shows that even epistemic grounds must be endowed with motivational force. Since facts or propositions lack motivational force, their obtaining is not sufficient to generate belief-revision processes.

12 IS PERCEPTION A SOURCE OF REASONS? 33 Second, some philosophers have compared the phenomenal character of experience with the force of assertions (see Matthen, 2005, ch. 13; Siegel, 2010, ch. 2). Does this mean that our requirement of assertive force precludes the possibility of a zombie having perceptually justified beliefs? 14 Although I find the notion of a zombie suspect, my previous remarks can be accepted even by someone who finds it intelligible. After all, the notion of assertive force is so defined that it is possessed by beliefs. As Lyons (2009, p. 73) defines it, a zombie is a counterpart of a normal human that is doxastically identical to her. Thus, the idea of a zombie is compatible with (BP), i.e., the claim that only beliefs can evidentially justify other beliefs. If the zombie lacked assertive states, it would lack perceptual beliefs altogether and, as a result, would lack perceptual reasons. We may probably conceive a being which lacks qualia but not one whose epistemically relevant states lack assertive force. 15 Third, by saying that assertive states are necessary we are not holding any internalist picture of justification. Recall that the internalism/externalism debate is related to the factors that determine the normative standing of propositions or beliefs, and whether they are (mostly) internal or external. By claiming that rational moves require an assertive dimension we are just making a point concerning the conditions under which a belief is rationally formed or revised in the light of perception. What the tribunal scenario shows is precisely that these moves do not occur because the subject entertains a content. They occur because she takes it at face value. And this is compatible with an externalist analysis according to which these moves are deemed justified only if belief revision routines are reliable or exemplify some rational principles that are not accessible to the subject. Having clarified the main point of the argument let me state a second constraint: Constraint 2: Any epistemological theory of perception must introduce assertive states as justifying states. It can do it in two ways: either it identifies justifying states with beliefs, or it conceives of experiences as assertive states. I shall examine these two options in the next sections. For the time being, let us reply to some potential objections. 14 I am indebted here to a referee of this journal. 15 I think that phenomenal features like the feeling of presence also play an epistemic role in the epistemology of perception. Nevertheless, I do not need to argue for this specific claim in order to criticize the strong propositional view. If the reader finds that phenomenology is irrelevant, she may read assertive force in functional terms, i.e., as a disposition to take p for granted when one uses it as a starting point for reasoning.

13 34 SANTIAGO ECHEVERRI 4. Propositional Justification and the Nature of Reasons In the last section, I suggested that my two arguments were neutral on the nature of reasons and justification. One might be inclined, however, to resist this idea. In this section, I will respond to some potential objections. 1. It might be objected that the previous objections overlook the orthodox distinction between propositional and doxastic justification. Intuitively, if p is justified for Pierre, Pierre is in a position to justifiedly believe that p, even if he fails to form the relevant belief (Kvanvig, 2003; Pollock and Cruz, 1999, pp ; see Turri, 2010, for criticism). If perceptual states must have doxastic force, there is no room for perceptual propositional justification. But, since the notion of propositional justification is fairly intuitive, something has gone wrong in the previous arguments. My reply is that propositions (or facts) themselves cannot ground beliefs. To be sure, there is a sense in which the fact that it is raining might be necessary for me to be justified in believing that it is raining. Nevertheless, this is not the sense intended by (SR), which is a claim about the way perceptual states are to be conceived in order to appropriately ground perceptual beliefs. (SR) qualifies a perception-belief relation, while the propositional notion of justification may hold even if no such relation obtains. When theorists like McDowell (1994a) and Brewer (1999) introduced (SR), they wanted to avoid Davidson s conception of mind as severed from the external world. The point of the previous argument was that it is not sufficient for perceptual states to have structured contents in order to appropriately link perceptual beliefs to the world. They must have assertive force. If they lack such force, they are inadequate to motivate rational moves like those involved in belief revision in the light of experience. 2. One might be inclined to resist the previous response by drawing a distinction between propositional and factive accounts of the contents of perception. According to some forms of disjunctivism, genuine perception is a relation to a fact. If the content of perception is a fact, one does not need to conceive of perception as having assertive force. After all, the content of a factive state is: it is the case that p. Thus, if one is perceptually related to a fact, one is given a 100% guarantee that it obtains. Hence, the factive view does more than just telling me that p, for it presents me with the fact that p. 16 The analogy between saying and experiencing can be understood in two ways. It may hold between the act of saying and the mode of experiencing or between the content of the speech act and the content of experience. The protest exploits the content side of the analogy. If one equates experience to the content of an assertion, 16 I owe this point to a referee of this journal.

14 IS PERCEPTION A SOURCE OF REASONS? 35 it will leave open the possibility that it is not satisfied, for propositions can be true and can be false. This is certainly less than the factive account promises, which purports to relate the subject to a fact. Nevertheless, what the tribunal scenario exploits is the other side of the analogy, the relation between the act of asserting and the mode of experiencing. From this perspective, if a perceptual state lacks assertive force, it does less than a state that tells me that p. After all, one does not rationally form a belief that it is raining because one is perceptually entertaining the fact that it is raining. Entertaining a content is always compatible with one s suspending judgement or casting doubt on it. Another way of putting this point is to realize that disjunctivism is an account of the content of genuine perceptual states, not of their attitudinal side. Disjunctivism is mainly an account of perception seeking to avoid epistemic intermediaries between mind and world. What the basing problem concerns, however, is the relation between perception and belief. Disjunctivism per se is silent on this score. The tribunal scenario shows that, even if our perceptual relation to the world lacks epistemic intermediaries, it is not sufficient to explain why one forms an appropriate belief when one is entertaining a fact. In order to prove the contrary, one has to show that, at least in the perceptual case, entertaining p is sufficient to be moved to judge that p. But, unless facts have motivational force, how is this claim to be established? It might be complained that some ontological pictures of facts make room for this option, as the following quotation from McDowell suggests: The point of the idea of experience is that it is in experience that facts themselves come among the justifiers available to subjects (McDowell, 1998a, p. 430; see also McDowell, 1995; Turri, 2009, pp. 501ff.). This sort of picture seems to be supported on introspection. Thus, Ginsborg (2006, p. 288) writes: What I consider when deciding what to do, or what it is rational to believe, is not my beliefs themselves but how things actually are. This verdict fits Dennis Stampe s (1987, p. 337) analysis: When we reason from our beliefs it is from what we believe the objects of our beliefs that we reason: the facts as we believe them to be. As I understand it, the argument is based on a model of deliberation as taking place in inner speech. On this view, when I am deciding what to do, expressions of the form I believe that never occur in the stream of consciousness. One does not reason that the streets are wet because one believes that it has rained. Instead, one thinks that the streets are wet because it has rained, period. In other words, one is typically aware of tokens of mental sentences like the streets are wet because it has rained, where there is no explicit talk about beliefs. 17 For further details on the relation between disjunctivism and conceptualism, see Echeverri (2011, section 7).

15 36 SANTIAGO ECHEVERRI This argument presupposes that, for a psychological state to be instantiated, there must be a mental word or concept that stands for it. But this assumption is preposterous. After all, a psychological state can be psychologically realized in an implicit way. In functionalist accounts of the mind, for example, the nature of belief states is defined by their functional relation to perception, other states, and behaviour. Similarly, an occurrent belief state can be instantiated when the mind performs the right actions, as when one asserts that it has rained or behaves as if it had rained. This shows that introspective data are misleading. They only work against pictures that equate belief states with occurrent (mental or public) utterances in which a mental term or concept stands for belief. But this picture is implausible. 18,19 4. Defenders of facts have to show that facts themselves, independently of any doxastic ingredient, can ground other beliefs. They might do it by exploiting the usual claim that facts can figure in causal explanations. In what follows, I shall respond to this line of argument. I will try to show that the causal efficacy of facts is not sufficient to rationally ground perceptual beliefs. More importantly, one can rephrase the main line of argument even if one decides to equate reasons with facts. Before we reach this conclusion, however, we have to examine the causal efficacy of facts. Consider these two examples: (1) The fact that Napoleon recognized the danger to his left flank caused him to move his troops forward. (2) It was the fact that the train was diverted that made me late for the lecture. 20 These fact-involving constructions cannot be seen as competitors of the doxastic analysis. Although the word fact is used in the two cases to introduce causal relations, it does not show that worldly facts, by themselves, rationalize belief revision processes. Example (1) owes its plausibility to the cognitive act introduced within the scope of the fact that : Napoleon s recognizing the danger to his left flank. If, however, one drops this ingredient, the explanation is defective. It could be that there was a danger to his left flank and Napoleon did not notice it. As a result, he would not have reacted in the appropriate way. Hence, it is not the worldly fact that there was a danger to his left flank that led him to react in the appropriate 18 The inadequacy of this picture should be obvious. It predicts that, in order to believe that p one has to possess the concept of belief. It would be like claiming that, in order to be hungry one has to possess the concept of hunger. 19 The introspective argument cannot be invoked to claim that reasons are facts. After all, belief-free or fact-involving utterances can be instantiated in one s stream of consciousness in false mental sentences. I may reason that it rained because the mental sentence the streets are wet was tokened in my mental fore. But it might occur that I misperceived the streets and, as a result, it was not a fact that the streets were wet. Thus, the introspective argument would lend support to the factive view only if introspected fact-involving constructions were immune to error. 20 I borrow both examples from Dodd (2000, pp. 91, 110 n. 4). He furnishes a number of references on the causal relevance of facts.

16 IS PERCEPTION A SOURCE OF REASONS? 37 way but his recognizing it. If we want to resist the Cartesian connotation of this remark, we can say that the fact should be coupled to his belief system in an adequate way. Napoleon should be poised to exploit that fact to motivate his behaviour. After all, there are many facts in the world that have no epistemic impact in one s cognitive life. Example (2) provides a case where facts can be used to introduce causal explanations. The trouble here is that this causal relation is not of the kind one is interested in when one claims that experience is a source of reasons. (SR) is not merely the claim that reasons arise from causal transactions with the world but the bold idea that perception can rationally ground belief. To make sense of the tribunal of experience, one needs a notion of causation capable of rationalizing belief revision processes. This is not what example (2) provides. There is no rational relation between the fact that the train was diverted and the fact that the speaker arrived late. The fact that the train was diverted does not figure as a reason in the subject s cognitive life. The justification is based on the assertion of the whole (2), and an assertion expresses a belief. 21 This discussion clearly shows that the arguments above do not depend on overarching premises on the nature of reasons. Even if one endorses the ontology of reasons as facts, one should supplement this picture with an account of the way these entities are present in one s cognitive life, eventuating in belief formation and revision. It is immaterial whether speakers introduce causal relations in factinvolving constructions. It is also irrelevant whether agents are unaware of forming sophisticated thoughts about their belief states. To decide whether a content can ground a belief, we have to consider the subject s attitude to that content. Bearing an attitude towards a content may be just treating it as true or being prepared to act as if it were true. In ordinary life, whenever I give a reason and I am being sincere, I do not give an unendorsed content. By the same token, if I decide to go and play tennis on the ground that the sun is shining, my behaviour is not rationally based just because the sun is shining. It is necessary that I take the content of my experience at face value. 22 We can summarize the first argument in the terminology used above. As I read him, Davidson s reliance on (BP) is based on the structural claim that sensations are not the sorts of things that can participate in justification relations. In my view, one can grant Davidson s point even if one has an ontology of reasons as facts, and denies Davidson s view of perception as bare sensation. In order to get the 21 There are knowledge accounts of assertion as well. Although I cannot discuss them here, let me make two points. First, if an assertion expresses a state of knowledge that p and knowledge entails belief, the argument presented here is vindicated. Second, knowledge accounts of assertion do not provide constitutive conditions of assertion, i.e., conditions without which assertions could not be made. After all, one can make false assertions. In those cases, one expresses a belief, not a state of knowledge. If the knowledge account were plausible, it would merely provide a regulative norm of the practice of assertion. 22 See Williams (2006, pp ) and Ernst (2001).